Loading ...
Sorry, an error occurred while loading the content.

4756Re: Nanosystems which imitate neural networks to realize a Nanofactor

Expand Messages
  • zander_cn
    Jan 1, 2005
    • 0 Attachment
      Thank you for the tip, Biodun (otonetafrica2000)

      I've added two chapters at
      http://www.01nn.com/nano/nano.htm

      1.)
      One local teacher for each nano unit versus
      one global teacher for many nano units

      Both methods have advantages and disadvantages.
      A global, external teacher wouldn't "know" the local conditions of
      each nano unit. However, a local teacher could optimally tune the
      relations between the nano unit and the environment.
      On the other hand, a global, external teacher would allow a massively
      parallel process. The only teacher could be e.g. an external computer.
      Perhaps in some cases both methods can be combined.


      2.)
      Nano units with Associative Memory

      Nano units with an Associative Memory based neural network inside
      could hypothetically give us the possibility to use only one global,
      external teacher for many nano units.
      All the nano units would learn synchronously and massively parallel.

      For example, the nano units have to learn:

      If input 1 then output x
      If input 2 then output y
      If input 3 then output z

      Neural networks with Associative Memory can learn this in the
      following way:

      The external teacher sends the associated signals to all the nano
      units always at the same time.

      The signal 1 and the signal x are externally(!) and globally(!)
      created at the same time.
      And so signal 2 and signal y.
      And so signal 3 and signal z.
      And so on.

      Carsten


      --- In nanotech@yahoogroups.com, biodun olusesi
      <otonetafrica2000@y...> > Will it not be more elegant to simply
      discard the other elements,allow the 'teacher' the sensing and
      reporting function under the control of an external computer? -
      something akin to what Robert Freitas Jnr. had written about! That way
      you have less space and less design headache. This comment however
      should be interpreted with respect to the theoretical feasibility of
      getting nanobots into the intracellular millieu for the purpose of
      subcellular imaging and manipulation.
      > Again, I personally see nothing wrong in borrowing of concepts,
      whether from AI, or any other source to illuminate how nanotech can
      move forward, and wish to commend you for your theory
      > Biodun
      > http://nanotology.org
    • Show all 12 messages in this topic